Re: the necessity of describing responses in-band

Hi Ruben,

> On November 9, 2015 at 11:05 PM Ruben Verborgh <ruben.verborgh@ugent.be>
> wrote:
>
>
> Hi John,
>
> > All it misses are some triples to link the retrieved resource to
> > the items
> > in the collection.
> >
> > <> hydra:member </items/45158567#id> , </items/35235179#id> ,
> > </items/10268448#id> .
>
>
> Yes, but suppose there is then also
>
> </items/45158567#id> :x :y
>
> and even
>
> :y :z :q.
> q: :r :s.
>
> Are these part of the data?

Well certainly they are part of the response document if they are in it ;-)
The answer probably lies in the eye of the beholder.

> And if they are (not), how can I express that?
>
> This example shows that it is hard to put a border around a member.

Yes it is, however seems the use cases we talk about are how to partition an RDF
graph
into sections/fragments to make it easier to understand/consume.

>
> > Basically seems like you are searching for recommendations/principles for
> > publishing RDF data sets on the web.
>
> Not just datasets; mainly hypermedia responses, really (which may contain
> parts of datasets).

This depends on what you mean by "dataset".

>From RDF1.1 spec the definition of an RDF dataset implies that any quads
document
can be considered as an RDF dataset.

If we then need to say that a collection of quads documents can be considered as
a dataset, this
risks that we head down a semantic rabbit hole.

>
> > Unfortunately I think the ship has already sailed here in the
> > broader RDF/SPARQL user base
>
> I wouldn't say that–many ships are leaking, and people are looking around ;-)

LOL, so long as we are not sinking yet!

>
> > - relate a named graph to the containing document using rdfg:subGraphOf
> > property
>
> Yeah; but the problem remains that, strictly speaking
> <g> rdfg:subGraphOf <>.
> and
> <a> <b> <c> <g>.
> are talking about different <g>s, according to the RDF 1.1 spec.

The way I read it, would be more correct to say "could be talking about
different <g>s".

> So we're still looking for a mechanism to say:
> when I name a graph <g>, and use it in subject/object position,
> I really mean the same thing.

Isn't Hydra CG in a position to make such a formal restriction for a certain
context (publishing Linked Data on the web)?

>
> Best,
>
> Ruben

John

Received on Tuesday, 10 November 2015 08:14:05 UTC